Rumor: Intel Coffee Lake benchmarks spill onto the web

The rumor mongers over at Videocardz have gotten a hold of benchmark figures for Intel's upcoming eighth-generation Core processors. Previous rounds of the rumor mill have reported that Intel will formally announce eighth-gen desktop chips on October 5. Numerous sources have claimed that the next-gen chips will have increased core and/or thread counts compared to the Kaby Lake chips on shelves now.

The first set of figures were Cinebench scores obtained by Canadian tech journalis Karl Morin on an HP Omen machine. The computer in question was fitted with an Intel Core i7-8700K chip with six cores and 12 threads running at 3.7 GHz. The Cinebench single-thread score was 196 and the all-core number was 1230. For reference, we obtained a nearly identical single-threaded score of 197 when we tested the Core i7-7700K back in January. That four-core, eight-thread 7700K could only muster 998 when tugging on all available hardware threads. As a second point of comparison, we managed a one-thread score of 160 and a multi-thread score of 1248 on AMD's Editor's Choice Award-winning Ryzen 5 1600X. It looks like the Ryzen 7 1800X's multi-threaded figure of 1647 might remain the best among mainstream desktop chips for a while longer.

The second set of numbers popped up on the Geekbench database and also came from a Core i7-8700K. The chip in questioned managed a single-core score of 5773 and a multi-core score of 24,620. The current Core i7-7700K gets around 5725 in the single-threaded test and 18,800 in the multi-threaded version. The previously-mentioned Ryzen 5 1600X musters 4180 on one thread and 18,650 with all threads active. AMD's fastest mainstream desktop chip, the Ryzen 7 1800X, gets about 4240 in the single-threaded test and 21,800 with all 16 threads tapped.

We can't speak to the accuracy of these figures, of course, but if the Core i7-8700K is indeed the six-core, 12-thread chip the rumor mongers say it will be, single-threaded figures on par with the Core i7-7700K and multi-threaded scores about 50% faster make a lot of sense. As usual, though, rumors should be accompanied by a generous helping of salt.

Comments closed
    • HisDivineOrder
    • 2 years ago

    Here’s hoping that this makes AMD lower prices more, which will put pressure on Intel to lower prices, which will put pressure on AMD to…

    …just like competition is meant to work.

    Alas, I bet they stabilize and meet in a coffee shop somewhere to arrange pricing at a table located between the Coke/Pepsi table, the Apple/Google table, and the Big Cable Table.

      • Beahmont
      • 2 years ago

      So you want them both to go out of business? Because you’re talking about a profit trends towards zero scenario.

      No profits mean no one is going to finance the next gen of process manufacturing.

        • Airmantharp
        • 2 years ago

        Don’t forget about supply: as soon as supply starts drying up due to lower prices creating more demand, prices will stabilize or increase.

          • Beahmont
          • 2 years ago

          That’s not the scenario in question though. HDO is explicitly talking about a Profits trend toward Zero scenario. Also price and profit are two distinct things that while linked are not the same.

          No one is going to spend billions of dollars to go even on an investment. HDO is literally wishing for the perfect competition scenario where price trends ever downward. No one wins in such a scenario however. Not even the consumers.

            • K-L-Waster
            • 2 years ago

            [quote<]No one is going to spend billions of dollars to go even on an investment. [/quote<] I think I see the RAM manufacturers looking a little sheepish in the back of the crowd...

    • gmskking
    • 2 years ago

    Geekbench isn’t a very reliable benchmark.

    • Zizy
    • 2 years ago

    50% higher MT at same TDP is nice, I expected slightly worse clocks from these chips. i5s could be very nice parts, probably also steal some sales from the R5 1600(X) and regain title of the “CPU to buy”.
    But I suspect the main thing these chips will achieve is decrease sales of 2066 stuff further. “Gaming workstation” will get possible without those expensive chips.

    I wonder when next Zens come and how they turn up. They have ~20% single threaded deficit. SB-like jump would be needed to overcome it. Considering they are going to a supposedly higher-clocked process, it could be even possible to get quite close to that with a modest IPC improvement.

      • Chrispy_
      • 2 years ago

      [quote<]But I suspect the main thing these chips will achieve is decrease sales of 2066 stuff further. "Gaming workstation" will get possible without those expensive chips.[/quote<] Yeah, 2066 is enough of a mess with desperate little market demographic willing to be stung by Intel, 12-thread with high-clockspeed and high-IPC will probably relegate S2066 platforms back to professional Xeon workstations again.

    • Firestarter
    • 2 years ago

    I really wonder how well it matches up against the 6 core and 8 core Ryzen chips in multithreaded benchmarks. It seems pretty safe to assume it’ll be the chip to beat for almost all game benchmarks

      • MOSFET
      • 2 years ago

      It will be strange looking at the upcoming benchmark charts, and seeing more Hyperthreaded AMD processors than Intel.

    • ozzuneoj
    • 2 years ago

    Every time I see any single threaded performance comparisons between Ryzen and Intel, I wish that Ryzen could simply clock higher. Being able to overclock a 1600X to 4.5Ghz+ would make it a much more interesting CPU. As it is there’s a major compromise in single threaded performance, even compared with older Intel chips (in certain situations), and overclocking is so limited that it doesn’t do much to change the situation.

    Look what they can do on LN2:
    [url<]http://www.overclockers.com/forums/showthread.php/782457-Ryzen-1600X-LN2-results[/url<] I highly doubt there will be a massive clock speed improvement with the next Ryzen, but it would be nice. Intel is finally starting to hit a clock speed wall again. Ryzen's IPC is fairly competitive, its TDP is quite low, the IHS is soldered... the only missing piece is a chip that is actually capable of significant overclocks.

      • xigmatekdk
      • 2 years ago

      It needs more than clockspeed. The latency is too high and is even worse as the core counts increases (Threadripper). Pointless synthetic benchmarks such as Cinebench won’t even show you this drawback. Which is why despite 1800x having similar Cinebench score as the 6900k, it is slower in real world apps (even more in games).

      • Anonymous Coward
      • 2 years ago

      Be thankful that AMD had the good sense to choose achievable targets for Zen. Hope that they continue to show such wisdom.

    • ronch
    • 2 years ago

    Well, it’s interesting to have a glimpse at how the two 6-cores will fare against each other (I’m a fan of Monster Bug Wars lately, coincidentally, so I imagine these two slugging it out like crazy insects). While I never expected the 1600X to hold out in multi-thread against any 6-core Intel, those numbers are pleasantly surprising. Single threaded, no one expected AMD to suddenly edge out Intel there so that’s cool with me. If those pricing rumors are true though and the 8700K is gonna go for $420, I think most folks would be happy to take home a $230 1600X instead.

    Edit – 230 ÷ 420 = 54.7

    Good grief this is even better than the days when AMD cloned Intel chips and sold them for about 20% less! Granted the performance was pretty much identical but hey, they just cloned them while they spent really money designing Ryzen!

    • smilingcrow
    • 2 years ago

    [url<]https://twitter.com/ryanshrout/status/907280680207503361[/url<]

    • ermo
    • 2 years ago

    Fitting 6 high-performance SMT-capable cores topping out at an all-core turbo of 4.3 GHz @ 95W is nothing to sneeze at.

    I expect this to rock in e.g. Project CARS 2 with VR and -pthreads 4 (4 physics threads which translates to one thread per Seta Tyre Model instance at each corner) since the physics engine is both very CPU hungry and because VR requires minimal jitter in order to be comfortable.

    At high vehicle speeds on laser-scanned tracks, the physics engine needs to process a LOT of track physics-mesh data per second when calculating the contact patch forces that determine how the car reacts (not to mention all the other simulation stuff going on under the hood).

    Even if it offers up a pair of extra cores, I honestly don’t know if the R7 @ 4-ish GHz will yield as good an experience in pC2…

      • homerdog
      • 2 years ago

      The 8700K will undoubtedly be the ultimate gaming chip.

        • Airmantharp
        • 2 years ago

        At least until Intel puts eight cores in a consumer socket!

      • ikjadoon
      • 2 years ago

      95W is a marketing tactic. Intel (like AMD on the RX480) hasn’t felt bound by their marketed TDP for ages.

      At load, the i7-6700K draws 20% over its 91W TDP.

      [url<]http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation/6[/url<] They swung it back with the i7-7700K, but don't be surprised if 50% more cores gives you higher than 95W. Intel doesn't care about the marketing TDP number. They sure as hell didn't optimize that many more cores/threads with only 4 more watts. I fully expect Skylake-like TDP trickery. I say this as a happy Skylake customer.

        • HERETIC
        • 2 years ago

        Why is “Thermal Design Power”still being miss-used as power usage?????????????????
        TDP is for heatsink manufacturers to know how much waste to dissipate…………………..

    • jihadjoe
    • 2 years ago

    Now this is the sort of performance we expected from Skylake-X.

    They really should have just stuck with the good old ring bus and cache config for CPUs with ≤12 cores.

    • DPete27
    • 2 years ago

    Zero IPC gains….nuts.

      • blargh4
      • 2 years ago

      Given that Intel has given no reason to think Coffee Lake’s core microarchitecture has changed from Skylake, why would there be IPC gains?

        • not_a_gerbil
        • 2 years ago

        Maybe more better cache? 32k lvl 1 seems small for 2017.

      • srg86
      • 2 years ago

      About what I was expecting, the same cores, just more of them. Even I would agree that calling them 8th Gen is a bit of a stretch, assuming there aren’t some nice upgrades to the iGPU.

      This is mainly more cores, and more features in the chipset.

        • smilingcrow
        • 2 years ago

        Considering how minimal the gains have been for many generations now it seems churlish to argue against this being a new generation when it adds 50% more cores.

      • K-L-Waster
      • 2 years ago

      The first test does say it’s running at 3.7 GHz which seems a little low by Intel i7 standards. It may be an early sample.

        • magila
        • 2 years ago

        The base clock is irrelevant for single thread tests. Base clock is largely determined by how fast the cores can run when under a worst case thermal load (read: AVX2) while staying within the specified TDP. A base clock of 3.7GHz is about what you would expect for six cores and a 95W TDP.

        For single threaded tests TDP is a non-factor so the turbo clocks are what matters. The rumored single core turbo clock for the 8700K is 4.7GHz. For comparison, the single core turbo clock of the 7700K is 4.5GHz.

        Edit: I should note that Intel CPUs tend to fall back to their “all cores” turbo clock more often than you’d expect. The 8700K has an all cores turbo clock of 4.3GHz compared to 4.4GHz for the 7700K which lines up pretty much perfectly with the reported results.

          • Beahmont
          • 2 years ago

          The first set of numbers are not turbo boosted, at least according to people who are supposed to know these things. The reported Cinebench score is using a fixed clock speed of 3.7 Ghz. If this is correct, there are IPC improvements in Coffee Lake.

            • magila
            • 2 years ago

            Says who? Running fixed at the base clock would be highly unusual, you’d have to go out of your way to configure it to run like that.

          • Srsly_Bro
          • 2 years ago

          Why were you downvoted?

            • magila
            • 2 years ago

            Beat me. Apparently people here don’t understand how Intel’s frequency scaling works, which is a little surprising and disappointing.

          • K-L-Waster
          • 2 years ago

          I’m sure that’s how it will behave once it ships. But again, this is a pre-release part, so may not have all functionality enabled.

          Since the doc for the test lists 3.7 GHz, that’s all we really have to go on.

      • Srsly_Bro
      • 2 years ago

      Don’t comment on products if you don’t understand the products. Coffee Lake is 14nm++. It is a process improvement, not an improvement of the architecture. There was 0 IPC gain from SKL to KBL, and the same from KBL to CFL.

      14nm, 14nm+, and 14nm++

      “I will enlighten you and banish some of your ignorance”

      -Judge Joe Brown

        • DPete27
        • 2 years ago

        I’ll admit guilt. I didn’t look at clockspeed. I also mistook Coffee Lake for Cannonlake in thinking that it was 10nm.

        • ikjadoon
        • 2 years ago

        PAO. What happen to PAO?! We got rid of tick tock and Intel said it’s now PAO. We didn’t even get through one damn cycle of PAO.

        Process – Architecture – Optimization

        Skylake was Architecture. Kaby Lake was Optimization. CFL was supposed to be Process, not a “plus” on the previous process. Intel should quit setting expectations erratically. Just say Moore’s Law is hard and we’ll release whatever we feel like that year.

        So now it’s PAOOOOOOOO?

        Take back your snark. DPete27 was on Intel’s confusing timelines. You, on the other hand, fell for marketing. Lithography is hard, but communication is easier.

          • Srsly_Bro
          • 2 years ago

          PAO is dead. Long live PAOOOOOO.

          CFL was never about Arch improvement; it was about process improvement. The snark stands. I would link you articles, but your ignorance will not be banished today. Reading is hard.

      • Airmantharp
      • 2 years ago

      None promised, but more cores are always welcome so long as IPC and clocks do not suffer!

    • chuckula
    • 2 years ago

    Don’t worry about spills, [url=https://www.cleanipedia.com/gb/laundry/how-to-remove-coffee-stains<]here's a comprehensive guide to cleanup.[/url<]

      • MOSFET
      • 2 years ago

      That’s oddly similar [url=https://techreport.com/forums/viewtopic.php?f=29&t=119019&hilit=rk+9000+bath<]to a local boy's story.[/url<]

Pin It on Pinterest

Share This